smash_config.json

{
    "batcher": null,
    "cacher": "fora",
    "compiler": "torch_compile",
    "factorizer": "qkv_diffusers",
    "pruner": null,
    "quantizer": null,
    "fora_interval": 2,
    "fora_start_step": 2,
    "torch_compile_backend": "inductor",
    "torch_compile_dynamic": null,
    "torch_compile_fullgraph": true,
    "torch_compile_make_portable": false,
    "torch_compile_max_kv_cache_size": 400,
    "torch_compile_mode": "default",
    "torch_compile_seqlen_manual_cuda_graph": 100,
    "torch_compile_target": "model",
    "batch_size": 1,
    "device": "cuda",
    "save_fns": [
        "save_before_apply",
        "save_before_apply"
    ],
    "load_fns": [
        "diffusers"
    ],
    "reapply_after_load": {
        "factorizer": "qkv_diffusers",
        "pruner": null,
        "quantizer": null,
        "cacher": "fora",
        "compiler": "torch_compile",
        "batcher": null
    }
}